Nov 4, 2024 · 39m · news
Sam Altman: What Startups Will be Steamrolled by OpenAI & Where is Opportunity | E1223 · 20VC with Harry Stebbings
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this live 20VC AMA interview at OpenAI DevDay, host Harry Stebbings sits down with OpenAI CEO Sam Altman to discuss the strategic future of reasoning models, the evolution of AI agents, economic scaling realities, and crucial advice for startup founders navigating a rapidly advancing AI landscape.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 22.9% of the talking time here. How this is scored →
Harry as informed peer 4.9 Guest teaching 4.9 Guest disagreement 2.0 Harry pushing back 3.3
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 30:10 Rejecting industry analogies and $100B entry cost
Sam explicitly dismisses Harry's historical technology analogies as a bad habit and rejects Larry Ellison's $100 billion foundation model entry cost quote.
Hardest push from Harry ▶ 20:46 Challenging Sam on Keith Rabois's under-30 hiring ruleHarry directly confronts Sam with controversial venture capital doctrine from Keith Rabois and Peter Thiel regarding under-30 founders, forcing Sam to defend his executive hiring philosophy.
Biggest teaching moment ▶ 30:50 The transistor analogy for AI progressionSam educates Harry on why standard historical analogies like electricity or the internet fail for AI, providing a detailed breakdown of why the transistor is the far more accurate conceptual model.
Harry holds his own ▶ 29:46 Citing Larry Ellison's $100B foundation model estimateHarry demonstrates high domain knowledge by quoting Larry Ellison's specific baseline cost estimate to challenge Sam on market entry barriers.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Welcome to OpenAI DevDay with Sam Altman | 5 | 4 | 2 | 2 | Harry asks targeted questions regarding o1 reasoning models, no-code capabilities, and where OpenAI sits in the tech stack versus RAG applications. Sam reframes how founders should evaluate model trajectory rather than building tools to patch short-term model shortcomings. | |
| From Betting Against to Betting For Model Improvements | 4 | 5 | 1 | 3 | Harry references a previous interview meme about OpenAI steamrolling startups and asks where opportunity exists. Sam explains the historical mindset shift from 95% of founders betting against model improvement to now betting for it. | |
| Evaluating the Economic Value and Capital Expenditure of AI | 5 | 4 | 2 | 3 | Harry cites Masayoshi Son's statement about $9 trillion in value offsetting $9 trillion in capital expenditure. Sam pushes back gently on focusing on exact macro numbers, emphasizing the order of magnitude of economic value creation instead. | |
| The Coexistence of Open Source and Integrated APIs | 4 | 6 | 2 | 2 | Harry probes on the definition and common misconceptions around AI agents. Sam reframes public perception, contrasting low-value tasks like making restaurant reservations with massively parallel workflows and smart senior co-workers. | |
| SaaS Pricing Models and the Compute-Based Economy | 6 | 6 | 3 | 5 | Harry presses Sam on model commoditization and whether models are depreciating assets given rising capital intensity. Sam bluntly rejects the premise that models aren't worth their training cost, explaining how revenue amortizes across ChatGPT's massive user base. | |
| Scaling Multimodality with Advanced Reasoning | 5 | 5 | 1 | 3 | Harry asks about multimodality scaling, RL paradigms, and life after transformers. Sam details OpenAI's core strength in pioneering unproven research paths rather than copying existing paradigms. | |
| Unlocking Wasted Human Potential | 4 | 4 | 1 | 2 | Harry asks about wasted human potential and how Sam's leadership style evolved over a decade of hypergrowth. Sam discusses the organizational difficulty of transitioning a company from 10% incremental growth to 10x step-function leaps. | |
| Balancing Youthful Audacity with Seasoned Experience | 6 | 6 | 4 | 5 | Harry challenges Sam with Keith Rabois and Peter Thiel's thesis that great companies must hire under-30 talent. Sam rejects the rigid age framing, explaining why massive compute infrastructure requires seasoned experts alongside young talent. | |
| Competitive Dynamics & System-Level AI | 5 | 5 | 2 | 4 | Harry brings up developer chatter about Anthropic models outperforming OpenAI at coding tasks and asks if scaling laws hit walls. Sam acknowledges Anthropic's performance while reframing the discussion toward system-level AI. | |
| Maintaining Team Morale and the Power of Shared Vision | 3 | 3 | 1 | 2 | Harry asks about team morale during failed training runs and how Sam manages 51/49 high-stakes decisions. Sam outlines his trusted network of domain experts rather than relying on a single advisor. | |
| Managing the Fractal Complexity of the AI Ecosystem | 7 | 7 | 4 | 6 | Harry cites Larry Ellison's claim that entering foundation model racing costs $100B and compares AI to the internet bubble. Sam forcefully rejects Ellison's figure and criticizes common historical analogies, offering the transistor as a far superior comparison. | |
| Quick-Fire Round: Tutors, Life-Context AI, and a Five-Year Vision | 5 | 4 | 1 | 3 | In a quick-fire round, Harry asks about vertical startup ideas, underrated research, and leadership weaknesses. Sam admits to feeling product strategy uncertainty, praising new hire Kevin Weil for bringing product discipline. |
Statements from this episode (29)
Altman: Startups patching current AI model flaws will become obsolete
“We are gonna try our hardest and believe we will succeed at making our models better and better and better, and if you are building a business that patches some current small shortcomings, if we do our job right, then that will not be as important in the futur…”
Altman: OpenAI o-series reasoning models will improve rapidly
“So you should expect Rapid improvement in the O series of models, and it's of great strategic importance to us.”
Altman: Building full startups with no-code AI will take a while
“It'll get there for sure. I think the first step will be tools that make people who know how to code well more productive. But eventually I think we can offer really high quality no code tools, and already there's some out there that makes sense, but you can't…”
Sam Altman: AI applications will create trillions in new market cap
“There will be many trillions of dollars of Market cap that gets created, new market cap that gets created by using AI to build products and services that were either impossible or quite impractical before”
Altman: 95% of founders now bet on rapid AI model improvements
“And so I felt like, 95% of people that were like, betting against the models getting better, five percent of people were betting for the models getting better. I think that's now reversed.”
Altman: OpenAI will make major push into next-gen AI systems in 2025
“Next year will be a big push for us into these next generation systems.”
Sam Altman: Open source AI models and integrated APIs will coexist
“There's clearly a really important place in the ecosystem for open source models. There's also Really good open source models that now exist. I think there's also a place for, like, nicely offered, well integrated services and APIs, and, you know, I think it's…”
Altman defines an AI agent as an autonomous long-duration task executor
“This is like my off the cuff answer. It's not well-considered, but something that I can give a Long duration tasks to, and provide minimal supervision during execution for.”
Altman: The most impactful AI agents will function like senior co-workers
“The category I think though is more interesting is not the one that people normally talk about where you have this thing calling restaurants for you, but something that's more like a really smart senior co-worker. Where you can, like, collaborate on a project …”
Altman: Enterprise software pricing could shift from per-seat to compute-based
“I'll speculate here for fun, but we really have no idea. I mean, I could imagine a world where you can say like, I want one GPU or 10 GPUs or a hundred GPUs to just be like churning on my problems all the time. And it's not like, you're not like paying per sea…”
Altman: OpenAI's o1 Model Demonstrates the Path to AI Agents
“There's a huge amount of infrastructure and scaffolding to build for sure, but I think O-one points the way to a model that is capable of doing great agentic tasks.”
Altman: OpenAI's model revenue justifies its high training costs
“This thing that they're not though worth as much as they cost to train, that seems totally wrong. To say nothing of the fact that there's like a, there's a positive compounding effect as you learn to train these models, you get better at training the next one,…”
Altman: Too many companies are training similar AI models without sticky products
“There's a lot of, there are probably too many people training very similar models, and if you're a little behind, or if you don't have a Product with the sort of normal rules of business that make that product sticky and valuable, then yeah, maybe you can't. M…”
Altman: Reasoning is OpenAI's most important area of focus
“Reasoning is our current most important area of focus.”
Altman expects rapid progress in image-based AI models
“Without spoiling anything, I would expect rapid progress in image. Based models.”
Altman: Competing AI labs will successfully replicate OpenAI's o1 model
“After, after a research lab does something, even if you don't know exactly how they did it, it's, I won't say easy, but it's doable to go off and copy it, and you can see this in the replications of GPT-IV, and I'm sure you'll see this in replications of O-one…”
Altman: Scaling 10x Requires Reinventing Operations, Not Repeating Past Methods
“One of the things that just came to mind out of, like, a rolling list of a hundred is how hard it is, or how much active work it takes to get the company to focus not on how you grow the next 10%, but the next 10 X. And growing the next 10%, it's the same thin…”
Sam Altman: Early-career talent shouldn't lead high-stakes infrastructure engineering
“When you're like designing some of the most complex and massively expensive Computer systems that humanity has ever built, actually like pieces of infrastructure of any sort, then I would not be comfortable taking a bet on someone who is just sort of like star…”
Altman: Anthropic has built an impressive AI model that excels at coding
“Yeah, they have a model that is great at coding for sure. And it's impressive work.”
Altman: Most software developers currently use multiple AI models
“I think developers use multiple models. Most of the time, and I'm not sure how that's all going to evolve as we head towards this more agentified world.”
Altman: AI industry focus will shift from models to systems over time
“Maybe if I had to describe it, we will shift from talking about models to talking about systems, but that'll take a while.”
Altman: OpenAI faced severe, unknown technical issues early in GPT-4 development
“Well, when we started working on GPT-IV, there were some issues that caused us a lot of consternation that we really didn't know how to solve. We figured it out, but there was definitely a time period where we just didn't know how we were gonna Do that model.”
Altman: Deep Belief in Deep Learning Pays Off Despite Major Setbacks
“There is something about betting on deep learning that feels like being on the side of the angels, and you kind of just, it eventually seems to work out, even though you hit some big stumbling blocks along the way, and so like a deep belief in that has been go…”
Altman: Leaders should consult specialized expert networks over single advisors
“No I think the wrong way to do that is to have one person you lean on for everything, and the right way to, at least for me, the right way to do it is to have, like, 15 or 20 people, each of which you have come to believe has good instincts and good context in…”
Altman: Unprecedented AI ecosystem complexity is his single biggest concern
“So, it's, supply chain makes it sign, sound too much like a pipeline, but yeah, the overall ecosystem complexity at every level of, like, the fractal scam is unlike anything I have seen in any industry before. And some version of that is probably my top worry.”
Altman: Entering the AI foundation model race will cost under $100 billion
“No, I think it will cost less than that”
Altman: Comparing AI to past tech revolutions is a bad habit
“Everybody likes to use previous examples of a technology revolution to talk about, to put a new one into more familiar context, and A, I think that's a bad habit on the whole, and, but I understand why people do it, and B, I think the ones people pick for Anal…”
Altman: The Cursor team has delivered a remarkable AI product experience
“Let me give a shout out to the cursor team. I mean, there's a lot of people doing incredible work in AI, but I think to really have, do what they'done and built, I thought about like a bunch of researchers I could name but in terms of using AI to deliver a rea…”
Altman: Product strategy is a personal weakness and current area of uncertainty
“The thing I'm struggling with most this week is I feel more uncertain than I have in the past about what our, like the details of what our product strategy should be. I think that product is a weakness of mine in general and it's something that right now the c…”
Shorts cut from this episode
▶ How will AI work FOR you? 💪 · 20VC with Harry Stebbings (@11:24)
▶ The most underrated AI company? 🔥 · 20VC with Harry Stebbin (@34:35)
▶ The best analogy for AI? 🤖 · 20VC with Harry Stebbings (@31:35)
▶ What would Sam Altman build if he started again? 🛠️ · 20VC (@32:52)
▶ What does Sam Altman think of Anthropic? · 20VC with Harry S (@23:42)
▶ Is your talent being wasted? · 20VC with Harry Stebbings (@17:22)
▶ What should AI startups build? 🤖 · 20VC with Harry Stebbing (@0:04)